-
Transaction
Paper
Introduction
Anglo Platinum’s Polokwane Smelter issituated outside Polokwane,
in the LimpopoProvince in South Africa. The plant wascommissioned
in March 2003. Wet concentrateis received from various
concentrators alongthe Eastern Bushveld Complex andoccasionally
from the Western BushveldComplex. On a daily basis, approximately
50%to 80% of the total concentrate received is fromthe UG2 reef and
50% to 20% is from thePlatreef and from the Merensky Reef.
Theconcentrate is dried in two flash driers andsmelted in a single
168 MVA furnace. Nominalthroughput of the furnace is 650 000 t/a
ofconcentrate at nominal power input of 68 MW.
In the furnace, the concentrate smelts intoa matte phase, a slag
phase, and some gas.Furnace slag is granulated, dewatered,
anddeposited onto a dump. A PGM-rich nickel-copper matte is cast,
crushed, and transportedto Rustenburg for converting. The process
hasbeen described in detail previously1.
Electricity is the single biggest input costfor smelting
platinum concentrate in an electricfurnace, currently exceeding 25%
of cash
operating cost. In combination with theelectricity shortage
experienced in South Africasince January 2008, and consistent with
aformal Anglo American drive to effect a 15%reduction in specific
energy consumptionacross all operations by 2014 (against anadjusted
2004 baseline), improving furnaceenergy efficiency is of paramount
importance.
Furnace design
The furnace at Polokwane Smelter (see Figure1) is a large
high-intensity furnace forsmelting platinum group metal
concentrates.The inside furnace dimensions are 29.2 m longand 10.1
m wide. The electrode diameter is 1.6m. The furnace has three matte
tapholes andthree slag tapholes located on opposite ends ofthe
furnace.
Furnace energy efficiency at PolokwaneSmelterby P.K. Van
Manen*
Synopsis
Anglo Platinum’s Polokwane Smelter is situated outside
Polokwane,in the Limpopo Province in South Africa. A single 168 MVA
six-in-line rectangular furnace smelts dry concentrates
containingplatinum group metals (PGMs).
Furnace energy consumption since commissioning is
presented,showing that energy efficiency improves as capacity
utilizationincreases. An energy balance for the furnace is
outlined, indicatingthe measured or calculated energy losses from
the hearth and side-wall cooling, the copper coolers, the upper
furnace walls and roof,and the energy losses in off-gas, matte, and
slag.
From the energy balance, the potential to improve
energyefficiency by controlling the slag temperature, the slag
level, andthe off-gas volume is derived. Furthermore, the impact on
theenergy balance of different concentrate types is discussed, as
wellas the potential impact of replacing all upper waffle coolers
by platecooler panels.
* Polokwane Smelter.© The Southern African Institute of Mining
and
Metallurgy, 2008. SA ISSN 0038–223X/3.00 +0.00. This paper was
first published at the SAIMMPlatinum Conference, Platinum in
Transformation,6–9 October 2008.
47The Journal of The Southern African Institute of Mining and
Metallurgy VOLUME 108 REFEREED PAPER JANUARY 2009 ▲
Figure 1—Polokwane Smelter furnace
74_van Malen:Template Journal 2/6/09 1:12 PM Page 47
-
Furnace energy efficiency at Polokwane Smelter
As a relatively high proportion of UG2 concentrate in thefurnace
feed was expected, the furnace was designed to treatconcentrates
containing up to 4% Cr2O3. Deep electrodeimmersion (up to 75%) and
high power intensity (up to 250kW/m2) should provide adequate
stirring to prevent hearthbuild-up and the formation of a
chrome-rich intermediatelayer between matte and slag.
Given the processing conditions described above, highermatte and
slag temperatures and higher sidewall energyfluxes were expected.
To ensure furnace sidewall integrity,Hatch waffle coolers were
installed in the slag and lowerconcentrate zone. The intense
cooling of the waffle coolersresults in the formation of a freeze
lining. The waffle coolershave been designed for a maximum energy
flux of 220kW/m2, but normal operating energy fluxes are 50–80kW/m2
with peaks of 110 kW/m2. The bottom and sidewallplates of the
refractory hearth are force-cooled by air.
The original Hatch waffle coolers unexpectedly showedsignificant
corrosion of copper at the slag/concentrateinterface. As a
consequence, the waffle coolers have had to bereplaced four times
since March 2003, while significantresearch and testwork was done
to solve this problem. At thefirst replacement, waffle coolers of
the original design wereinstalled because of the urgency to get the
furnace back inproduction. However, at the second replacement, the
designwas changed from a 1.2 m single-height waffle cooler to
twohalf-height coolers (lower and upper waffle coolers), with
theupper coolers being split again vertically in half. The
reasonfor this was the observation that the corrosion
occurredpredominantly on the upper half of the coolers at the
slag-concentrate interface.
Sulphidation (and even some chlorination) of copperappeared to
be the main corrosion mechanism. Differentmaterials were tested to
protect the upper waffle coolers, butthe installation of graphite
blocks on the hot face appeared tobe the most effective. Therefore,
graphite has now beeninstalled on all the upper waffle coolers.
Some of the upperwaffle coolers were replaced by a ‘plate cooler
panel’, as atest. The plate cooler panel is a staggered arrangement
ofthree copper plate coolers and three graphite plates.
Thecorrosion process and remedial action have been described
indetail elsewhere2.
This paper focuses on the furnace energy efficiency. Itshows the
historical energy consumption during the fivecampaigns (the period
between two waffle coolerreplacements is referred to as a campaign)
and the trend in(gross) specific energy consumption. An energy
balance ispresented based on actual measurements of the
energylosses, and the potential to improve energy efficiency
byreducing energy losses is discussed.
The specific energy consumption is also a function of
themineralogy of the concentrates smelted. Therefore, the impactof
different concentrate blends on the day-to-day specificenergy
consumption is evaluated. Finally, the potentialimprovement in
energy efficiency if the upper waffle coolerswere replaced by plate
cooler panels is shown.
Furnace energy balance
On an hourly basis, and even on a daily basis, the furnace
isseldom in true thermal equilibrium, as matte, slag, and bone-dry
levels and concentrate blends change continuously.Therefore, it was
decided to do monthly energy balances(Table I). The losses from the
furnace are shown schemat-ically in Figure 2.
The furnace power input is calculated by the processcontrol
system from the electrical parameters on the primaryside of the
transformers, giving a daily power input figure.
The energy required for smelting concentrate is calculated
▲
48 JANUARY 2009 VOLUME 108 REFEREED PAPER The Journal of The
Southern African Institute of Mining and Metallurgy
Table I
Analysis of monthly furnace energy losses for May and July
2008.
Design May 08 July 08
Energy output (MWh/h):Off-gas energy losses 4.96 4.5 3.4Dust in
off-gas 0.58 0.26 0.34Copper coolers 2.6 2.1Hearth cooling 1.2
1.2Sidewall cooling 0.3 0.3Cooling of LV buses and contact pads
0.43 0.53Roof energy losses 0.2 0.2Sidewall energy losses 0.11
0.11Transformer no-load losses 0.14 0.14Transformer energy losses
0.19 0.25Electrode energy losses -0.09 -0.11Total energy losses 10
9.84 8.46
Numbers are in MWh/h (or MW) unless stated otherwise
Table II
Average (net) specific energy requirement forsmelting Platreef,
Merensky, and UG2 concentrate,using July 2008 data
Concentrate type Specific energy requirement(kWh/t
concentrate)
Platreef 579Merensky 554UG2 66747/3/50 % mix of Platreef/Mer/UG2
62217/3/80 % mix of Platreef/Mer/UG2 649
74_van Malen:Template Journal 2/6/09 1:12 PM Page 48
-
by converting the monthly weighted average chemicalcomposition
of each concentrate type into a mineralogicalcomposition. The
enthalpy increase and enthalpy of fusionfor each sulphide and oxide
mineral is used to calculate howmuch energy is required for melting
each concentrate type.Table II gives an overview of the average
energy requirementfor the main types of concentrate smelted
(Platreef,Merensky, and UG2).
It can be seen that the specific energy requirement ispredicted
to increase by about 4% from a 50% UG2concentrate mix to an 80% UG2
concentrate mix. This changecan occur from one day to the next and
will affect the feedrate required for a fixed furnace power input.
However, thechanges occur gradually during the day and can be
managedeasily by hourly soundings and slag
temperaturemeasurements.
The energy losses consist of the following components:• Low
voltage bus and contact pad cooling—there is a
separate cooling water circuit for the low voltage busesand
contact pads. The energy loss was calculated fromthe water flow and
the temperature difference betweenthe supply and return cooling
water.
• Roof energy losses—these were calculated using thethermal
conductivity of the bricks (high densityalumino-silicate fire
brick) and the averagetemperature difference between the inside and
theoutside surface of the bricks (300ºC).
• Upper side-wall energy losses—these were alsocalculated using
the thermal conductivity of the bricksand the average temperature
difference between theinside and the outside surface of the bricks
(433ºC).
• Copper cooler heat losses—the waffle coolers are cooledby a
closed water circuit operating between 39ºC and41ºC. The energy
removed by the waffle coolers andplate coolers is estimated from
the rise in temperatureof the closed-circuit cooling water before
and afterpassing through the coolers.
• Sidewall cooling energy losses—these are calculated
byestimating the air flow from the fan curve, motorpower, and
pressure differential. With the ‘specificheat’ of air, and the
temperature difference betweeninlet and outlet, the energy loss can
then be calculated.
• Hearth cooling energy losses—the energy loss iscalculated in
the same way as for the sidewall cooling.
• Off-gas energy loss—this is calculated using an off-gasvolume
from the furnace of 20 000 Nm3/h at atemperature of 600ºC and 0.7%
SO2 and 3% moisture.
• Dust in off-gas energy loss—this was calculatedassuming 3% of
feed mass is entrained in the off-gasflow and leaves the furnace at
500ºC. The dust isreturned to the furnace at 25ºC and a ‘specific
energy’of 1 MJ/t/K has been used.
• Electrode energy losses—the solid electrode paste mustbe
heated, melted, baked, and further heated to at least1 600ºC. It
has been assumed that the fixed carbon inthe paste is oxidized to
carbon monoxide inside thefurnace and that the carbon monoxide is
heated to 1 700ºC at the electrode tip. The oxidation of the
carbonin the paste provides more energy than is required forheating
the paste and the casing steel, so that there isan energy gain. It
has also been assumed that the pasteheaters provide the energy for
melting the paste. Apaste consumption of 2.9 kg/MWh and a
steelconsumption of 0.3 kg/MWh has been used.
• Transformer losses—these consist of no-load losses(obtained
from the transformer manual) and energylosses dissipated to cooling
water, which are calculatedfrom the water flow and the temperature
differencebetween incoming and outgoing cooling water. Allwater
temperatures were measured by taking a 5 lwater sample and
measuring the temperature with amercury thermometer.
Comparing the May and July energy balances with thedesign, it
can be seen that the actual energy losses are asmuch as 15% lower
than the design heat losses. This was tobe expected for a
conservative design.
Historical furnace energy consumption
Figure 3 shows the historical energy consumption for each ofthe
five campaigns since January 2004. Although the furnacewas
commissioned in March 2003, only data from January2004 is taken
into account, as the furnace operated at verylow power input and
with lime addition during 2003, and theconcentrate mix smelted was
not comparable to later periods.Campaigns 1 to 4 show that specific
energy consumptionreduces as average furnace power increases. This
is to beexpected, as the energy losses do not increase
proportionallywith the increase in furnace power.
In order to show the impact of a period of operation athigher
furnace power, a data point for July 2008 was added,when the
average furnace power input for the month washigh, and gross
specific energy consumption was reduced by6% compared to the
average of campaigns 3 and 4. Thisclearly demonstrates that as the
furnace power increasesfrom 44 MW to the nominal design of 68 MW,
the energyefficiency improves less than when the furnace operates
atlower power. Campaign 5 is an outlier and no obvious reasonhas
been found for the lower specific energy consumption.
Potential to improve energy efficiency
A large furnace operates more energy-efficiently at higher
Furnace energy efficiency at Polokwane SmelterTransaction
Paper
49The Journal of The Southern African Institute of Mining and
Metallurgy VOLUME 108 REFEREED PAPER JANUARY 2009 ▲
Figure 2—Schematic representation of energy losses around
thefurnace
74_van Malen:Template Journal 2/6/09 1:12 PM Page 49
-
Furnace energy efficiency at Polokwane Smelter
smelting rates, as the energy losses do not increase
propor-tionally with the smelting rate. However, regardless of
thethroughput, energy efficiency can be improved by loweringthe
slag temperature and the slag and matte levels, and byreducing the
off-gas energy losses. The lower energy lossesshould result in a
lower electrical power input for the samesmelting rate or result in
a higher smelting rate for the sameelectrical power input, or a
combination of both.
Slag temperature
The slag temperature plays a role with the energy transferfrom
the slag bath to the ‘black top’ (unmelted concentratelayer on top
of the slag), and with the smelting rate ofconcentrate that is
stirred into the slag bath. Higher smeltingrates require higher
energy transfer rates and therefore athigher smelting rates the
slag temperature tends to be higher,especially if the concentrates
being smelted contain species ofhigh incipient fusion that can
adversely promote sintering inthe blacktop. At 65 MW, slag
temperatures as measured inthe slag launder with a handheld optical
pyrometer rangefrom 1 570ºC to 1 600ºC (although historically slag
tappingtemperatures have been measured as high as 1 700ºC). It
isassumed that the slag temperatures can be at least 50ºChigher
inside the furnace (although this remains to bequantified).
During conditions of underfeeding, the slag temperaturecan
increase by 20ºC within 30 minutes. Assuming aconcentrate feed mix
as for July 2008, the impact on thefurnace energy balance is
predicted as follows: At typical feed
rates a 20ºC increase in slag temperature would consume 0.5MW of
additional power, or 0.8% of the average electricalpower input, and
smelter production for the day would bereduced commensurately.
Slag level
After the furnace off-gas, the main source of energy loss isthat
from the copper coolers (about 25% of the measuredenergy losses),
which are at the slag level. However, if thematte level is lower,
less area of copper cooler is covered byslag, so both slag and
matte levels affect the energy lossesthrough the copper coolers.
The energy loss through thecopper coolers was measured at ‘high’
slag levels and at ‘low’slag levels (see Table III).
The data suggest that 0.4 MW would be saved for every10 cm
decrease in slag level, or, alternatively, if the slag levelwould
be operated 10 cm higher, 0.4 MW additional powerwould be required
for the same smelting rate, or lessconcentrate would be
smelted.
The ability to lower slag levels is limited because the
slag-concentrate interface must always be at the upper
wafflecoolers to ensure that the lower waffle coolers are
notunnecessarily exposed to corrosion.
Off-gas volume
The energy loss in the off-gas has been calculated assuminga gas
volume of 20 000 Nm3/h at 600ºC. The off-gas volumeis made up
predominantly of sulphur dioxide and moisturefrom the concentrate
(assume 0.5 mass %), air supplied
▲
50 JANUARY 2009 VOLUME 108 REFEREED PAPER The Journal of The
Southern African Institute of Mining and Metallurgy
Figure 3—Historical indexed specific energy consumption as a
function of furnace power input per campaign (data for campaign 1
is taken to represent100%)
Table III
Energy loss through copper coolers at high and low slag
levels
Slag level Total bath height Rate of energy loss Cooler area
covered by slag Difference(cm) (MW) (m2) (kW/m2)
High 178 2.8 75.1Low 151 1.4 53.4 52.6
74_van Malen:Template Journal 2/6/09 1:12 PM Page 50
-
through the concentrate feed system and ingress air
(CO/CO2combustion products of the electrodes have been ignored
forthe purposes of this analysis—Table IV).
It can be seen that most of the off-gas volume consists
ofingress air. Therefore, the off-gas volume can be reduced
byimproved sealing of the furnace. A reduction in off-gasvolume of
10% would make 0.44 MW more available forsmelting, and achieve an
equivalently higher amount ofconcentrate smelted.
Slag temperature, slag level and off-gas volume eachallow
relatively small changes. However, together these smallimprovements
could add at least 2% to the smelting rate,which becomes
significant from an economic point of view ifthe furnace is the
bottleneck, and when optimum energyefficiency is a driving force
for smelter improvement.
The impact of plate cooler panels
As part of the investigations to mitigate the corrosion of
thecopper coolers, some upper waffle coolers were replaced by
a‘plate cooler test panel’. The plate cooler test panel consists
of
a staggered arrangement of three copper plate coolers andthree
graphite plates. There are two upper waffle coolers(UWC) for one
lower waffle cooler (LWC) and two sets ofplate coolers for one
lower waffle cooler. Plate coolers wereinstalled above lower waffle
coolers 39 and 40. Lower wafflecoolers 38 and 41 had one set of
plate coolers and one upperwaffle cooler installed on top of them.
Lower waffle coolers37 and 42 had normal upper waffle coolers
installed abovethem (see Figure 4). Therefore, the plate coolers
above lowerwaffle coolers 39 and 40 are compared to the upper
wafflecoolers above lower waffle coolers 37 and 42.
Compared to the normal upper waffle coolers, only halfthe
surface area of the hot face consists of water-cooledcopper.
Therefore, it is to be expected that the energyremoved from the
plate cooler test panel is less than theenergy removed from a
normal upper waffle cooler. Thedifference in energy flux is
estimated using historical data ofcooling water temperature
differences from September 2007.The reason for using historical
data is that the plate coolertest panels were not reinstalled after
the run-out in February,as this would have delayed the restart of
the furnace.
When the slag levels are low, only a small part of theupper
waffle coolers or plate coolers is covered by slag.Therefore, the
difference in energy loss between the upperwaffle coolers and the
plate coolers is postulated to beinsignificant when slag levels are
low.
Data have been collected when cooling water temper-atures, and
thus slag levels, were relatively high, and theenergy removed by
plate coolers versus upper waffle coolershas been calculated using
temperature differences betweeninlet and outlet cooling water for
each cooler. Figures 5 and 6show the detail arrangements of coolers
and water circuits for
Furnace energy efficiency at Polokwane SmelterTransaction
Paper
The Journal of The Southern African Institute of Mining and
Metallurgy VOLUME 108 REFEREED PAPER JANUARY 2009 51 ▲
Table IV
Breakdown of off-gas volumetric rate
Source Volume (Nm3/h)
Sulphur dioxide 356Moisture 605Feed system 4 300Ingress air 14
739Total 20 000
Figure 5—Arrangement of two upper waffle coolers and water
circuits
Figure 4—Arrangement of coolers around plate cooler test
panel
74_van Malen:Template Journal 2/6/09 1:12 PM Page 51
-
Furnace energy efficiency at Polokwane Smelter
upper waffle coolers and plate coolers. The energy loss hasbeen
calculated for a set of plate coolers and for a set of upperwaffle
coolers above lower waffle coolers 37, 39, 40, and 42,assuming that
all coolers are exposed to the same conditionsat the same time
(Table V).
The energy removed above LWC 40 is not in line with theother
data, and it is suspected that there is a problem withone of the
water temperature measurements. Therefore thesedata are discarded.
At high slag levels, there is a 7.1 kWdifference between the rate
of energy removed from the platecoolers above LWC 39 and the
average rate of energy transferfrom the upper waffle coolers above
lower waffle coolers 37and 42. The area of one set of plate coolers
above a lowerwaffle cooler is 1.138 m2. The total area of upper
wafflecoolers is 50.69 m2. For the whole furnace the reduction
inenergy losses is estimated to be 0.3 MW, if slag levels arehigh,
which occurs typically less than half the total operatingtime.
Since the requirement is to keep slag levels low to reducethe
loss of energy through the copper coolers, and since thedifference
in energy removed by plate coolers and upperwaffle coolers is
small, it would not be worthwhile replacingthe upper waffle coolers
by plate coolers from an energysavings perspective. It might make
more sense to cover thelower waffle coolers with graphite as well,
allowing the
furnace to operate at a lower slag level, provided that this
hasno adverse impact on electrode proximity to the matte, andmatte
temperatures.
Conclusions
The energy balance shows that actual energy losses are lowerthan
design energy losses, so that the furnace should be ableto achieve
design throughput.
Historical energy consumption by campaign, excludingfurnace
heat-up, shows that specific energy consumptiondecreases as average
furnace power increases, up to inputlevels of 59 MW.
The smelting rate could be increased by at least 2% torealize
specific energy savings by further optimizing slagtemperature,
total bath level, and off-gas volume.
It is not recommended that upper waffle coolers arereplaced by
plate coolers, from an energy savings perspective.
References
1. HUNDERMARK, R., DE VILLIERS, L.P.S., and NDLOVU, J. Process
description andshort history of Polokwane Smelter, Southern African
Pyrometallurgy2006, R.T. Jones (ed.), SAIMM, Johannesburg, 5–8
March 2006, pp.35–41.
2. DE VILLIERS, L. P. S., HUNDERMARK, R., NELSON, L.R., and
VIVIERS, P. Privatecommunication, 2007.
3. NELSON, L.R., STOBER, F., NDLOVU, J., DE VILLIERS, L.P.S.,
and WANBLAD, D.Role of technical innovation on production delivery
at the PolokwaneSmelter, Nickel and Cobalt 2005: Challenges in
Extraction andProduction, 44th Annual Conference of Metallurgists,
Calgary, Alberta,Canada, 21–24 August 2005, pp. 91–116.
4. YUHUA PAN, SHOUYI SUN, and SHARIF JAHANSHAHI, Efficient and
PortableMathematical Models for Simulating Heat Transfer in
Electric Furnaces forSulphide Smelting, International Symposium on
Sulphide Smelting 2006,F. Kongoli and R. G. Reddy (eds.), The
Minerals, Metals and MaterialsSociety.
5. CRC HANDBOOK OF CHEMISTRY AND PHYSICS, CRC Press, 63rd
edition,1982–1983.
6. ROSENQUIST, T. Principles of Extractive Metallurgy,
McGraw-Hill BookCompany, 2nd edition, 1983. ◆
▲
52 JANUARY 2009 VOLUME 108 REFEREED PAPER The Journal of The
Southern African Institute of Mining and Metallurgy
Figure 6—Arrangement of one plate cooler test panel and water
circuits
Table V
Energy removed by plate cooler test panels andupper waffle
coolers at low and high slag levels
Plate or UWC LWC number Energy removedLow slag level High slag
level
(kW) (kW)
UWC 37 30.5 52.0Plate cooler 39 38.5 44.1Plate cooler 40 10.5
20.5UWC 42 29.6 50.5
74_van Malen:Template Journal 2/6/09 1:12 PM Page 52